Support for custom models

[danielrosehill]

Hey!

Recording this using Speech Notes :-).

I recently created a fine tune on Whisper and downloaded the weights for local use as a .safetensor file.

I know there is the ability to define a custom storage path from models. But I was wondering if there is any native way to point Speech Note to a custom model for inference.

If there is not, it would be a really cool feature to integrate, as the process wasn't too difficult, but I've definitely noticed a step up in accuracy.

For what it's worth, the fine-tune was based on the Whisper Large V3 Turbo.